correcting for test type proportion using empirical proportion observations in the data can lead to volatility in days of anomalous proportions. Suggested improvement is to have a time-constant proportion multiplier for one test type over another, and have it learned through fitting. Need to think about this more to avoid identifiability issues
correcting for test type proportion using empirical proportion observations in the data can lead to volatility in days of anomalous proportions. Suggested improvement is to have a time-constant proportion multiplier for one test type over another, and have it learned through fitting. Need to think about this more to avoid identifiability issues